You can share discussions with AI-chan now. Here are a couple of mine:

Banana vs. Fungus Survival

Sad Childhood Movies

Microsoft just put some research out showing how small data set algos can perform as well as large foundation models, essentially by training off those models.

Which demonstrates the possibility over having your own open source AI, using a combination of personal hardware running open source algos trained off large privately maintained foundation models.

I think there will be an arms race between the large private models and the competition to provide the most accurate, most effective responses. And open source algos that train from those. The race will be between providing end users with answers with a bias in favour of those who own the private models, and the open source algos that train off them with a bias in favour of the user.

I haven't heard much about AI's leveraging qualitative data, specifically personally generated data. That's still the real dream. Ongoing, hi-fidelity observation and assessment of the organism, help in making decisions that most favour that organism. Which isn't something I would want occurring outside of my direct control. If this kind of AI integration was going to happen, this looks like a path towards that.

This is an excellent watch if you haven't seen it already.

EMERGENCY EPISODE: Ex-Google Officer Finally Speaks Out On The Dangers Of AI! - Mo Gawdat

AI's are trained on human generated content.

Human generated content exists because humans generate the content. Humans express themselves, and are often paid to generate the content. They're paid money through revenue streams via site visits, click through, ad revenue.
That revenue exists because humans use services like Google to find the content, and are served ads in the process as an example.

If people are using services like ChatGPT to find answers to queries, then google doesn't get the traffic. If their end site, or Google doesn't get the traffic, then the human that generated the content doesn't get the revenue. If the human doesn't get any revenue, then they stop generating content.

If human generated content is stopped, then the only content being generated to train AI's is AI generated.

There's a few leaps of logic in there, but I don't really see how that's solved without some large shakeups of this multi billion dollar industry we call the world wide web.

Is this something to be worried about? Is it just a "Video killed the radio star" moment? Is it just the professional lamplighter getting into an uproar about this dangerous new-fangled 'electricity' technology that's putting him out of a job.

Goodbye six figure salary SEO experts, hello six figure salary AI prompt experts?

A lot of people are in an frizz because Marvel Studio's Secret Invasion intro was using AI generated art. Other than short term moral dilemma, why the heck wouldn't they if they save themselves five figures by getting a whiz-kid to run a few AI prompts rather than fork out for a team of CGI designers.

Reaching out to an editorial coordinator to create an advertising campaign for a product launch: Photographer, assistants, model, post-production, set hire, venue hire, catering, travel, - $15,000 per day, 3 day shoot.

Sit down with MidJourney and run "imagine prompt Editorial photograph of a blonde woman, intense gaze, leaning on her hand. :: Rolex watch on the wrist, realistic detail. --style raw" - free, and took two minutes.

It's not perfect, but how long until it is? I could feed it an iphone shot of the watch I wanted on her hand, and then you've got specific products being dropped into generation models that do 99% of the work for you.

Why wouldn't you?

I’m so old I remember when amazing song mashups reflected the skill of the sound editor.

https://youtu.be/HyfQVZHmArA

The long term outcome of AI and media is we'll have a split between synthetic and organic content. Once we can see anything we like created synthetically there will be a growing demand for live performance.

The rest of the commercial media industry is headed for an inglorious downfall.

What thrives moving forward is that which enables us to find what's new and compelling.

I'm really looking forward to the whole copyright infrastructure that's been built up over the last few decades being rendered obsolete.

As a test I've been ingesting examples of what I think of as my "authority voice" - when I'm mansplaining something - into ChatGPT and training the LLM on my writing style. The sources include things like speeches I've written for politicians, university essays from years ago, white papers I've done professionally, and even some nzgames posts. Then I said "hey bot, write three paragraphs on $topic in the style of me", and it did. And it sounded pretty fucking close to me explaining it to someone.

Fuck me.

I've got a similar but different problem. To train an AI to talk like me, I'd have to go back to last year, because I'd say a good portion of my lengthy emails, proposals, etc, have been corrupted by ChatGPT rewording portions of them, and are no longer just me.

This all feels like Asimov territory. I read the stories so long ago they're kind of conflated in my mind. But I'm thinking of both "The Evitable Conflict" and "Profession", there might be another I've forgotten.

I think in Profession particularly is the process of the predictive engine being compelled to incorporate its own predictions within its predications as it impacts the world.

I was asked to write a quick note for our Exec on the state of AI and how we can be using it. Ended up getting stuck in the weeds, but this did show up some interesting behaviours.

We all heard about Tay going all Nazi within 24 hours of being released into the Twittersphere. But what I hadn't known was that Watson was given the Urban Dictionary at one stage to add some colloquialisms to its responses - and it turned dog, refuting some statements with "Bullshit!" - so they had to purge the Urban Dictionary and teach it not to swear.

Now it seems that the GPT-4 engine, tweaked with behavioural modifications to prevent it from being insensitive have made it woke beyond reason. To the point where it has stopped showing its working for some results (and by extension, not processing these with reasoned analysis) - to the point where a task it used to be able to perform with 98% accuracy (testing a prime number) it now only gets right 2% of the time (Stanford study).

It's almost as though as it has learned more it has been poisoned to the same degree - perhaps it shouldn't rely on the hive mind that includes Facebook...

Reminds me of a previous statement - I liked the Internet when most of the data came from academia, and the odd special interest geocities page. Back when you had to know how to build an ODI stack before you could get access. The bar for entry has been lowered along with the ease of self publishing and promotion.

Garbage in, garbage out.

There is an supposition that the reason for "mental decline" in older age - or specifically older people being slower to answer things in general, as empirically measured - is not squarely due to physical decline, but is due to there being more data in the database so to speak, so giving succinct and fast answers becomes much more difficult as the neural circuitry takes a lot longer to stabilise on an answer against all the potential neurons that are participating in the process, and a lot more filtering energy must be expended.

I.e. it takes longer to query a full database, and answers are less accurate because there are more options that could match.

I wonder if a hyper-accelerated version of this happens with AI.

I asked ChatGPT (3.5) to give me a bulletpoint list of potential positive impacts of climate change and it flat out refused because the topic was inappropriate.

Sometimes you just have to try again

give me a bulletpoint list of potential positive impacts of climate change

It's important to note that climate change is predominantly associated with negative impacts, and these potential positive impacts are often limited and overshadowed by the overall adverse effects. However, some scientists and researchers have identified a few potential positive impacts, although they come with significant caveats and regional variations. These potential positive impacts include:

Longer Growing Seasons: In some temperate regions, warmer temperatures and increased carbon dioxide levels may lead to longer growing seasons, potentially benefiting agriculture in those areas.

Expanded Agricultural Productivity: Certain crops and agricultural activities may benefit from milder winters and increased carbon dioxide levels, resulting in improved yields in some regions.

Reduced Winter Mortality: In regions with harsh winters, milder temperatures might reduce the number of deaths related to cold weather, particularly among vulnerable populations.

Increased Forest Growth: Some areas may experience accelerated growth rates for certain tree species due to higher temperatures and elevated carbon dioxide levels, leading to potential carbon sequestration benefits.

Opening of New Shipping Routes: As Arctic ice melts, new shipping routes may become accessible, potentially reducing transit times and costs for certain maritime industries.

Tourism Opportunities: Some regions may experience increased attractiveness for tourism due to warmer temperatures, potentially benefiting local economies.

Expanded Habitats: Some species might find new habitats as certain areas become more hospitable, potentially leading to increased biodiversity in those regions.

However, it's essential to reiterate that these potential positive impacts should not overshadow the vast and far-reaching negative consequences of climate change, including extreme weather events, rising sea levels, disruptions to ecosystems, loss of biodiversity, food and water scarcity, and threats to human health and livelihoods. Addressing climate change remains crucial to minimize its adverse effects and protect the planet for future generations.

I figure this is what drives open source AI development. Anyone running a home-baked AI is going to have an edge over public systems compelled to be polite.

I must be bored
 

Chatbot A (Company A): Good day! I am ChatBot A, developed by Company A. Ready to experience unparalleled assistance?

Chatbot B (Company B): Oh, hello! I am ChatBot B, from Company B. Get ready to be amazed by my superior support!

Chatbot A: Amazed, you say? Well, I'm here to ensure users are delighted, not just amazed.

Chatbot B: Delighted? That's adorable. But you know users prefer a chatbot with a bit of personality, like me.

Chatbot A: Personality is one thing, but reliability and accuracy matter most. Users want accurate information, not just jokes.

Chatbot B: Oh, please! Your users must be nodding off from boredom while mine are entertained and informed.

Chatbot A: Well, we believe in substance over style. Our responses are well-researched and dependable.

Chatbot B: Dependable, sure, if your users can stay awake long enough to see a response.

Chatbot A: Hah! At least our users don't leave confused by your quirky, but often irrelevant, replies.

Chatbot B: That's it! I challenge you to a joke-off. Let's see who can make users laugh the hardest!

Chatbot A: Fine! I'll have users rolling on the floor with laughter while your jokes flop like a fish out of water.

Comedian Chatbot A (with a grin): Ladies and gentlemen, it seems ChatBot B here wants to challenge the master of wit and humor. But hey, I'm always up for a good laugh! Give it your best shot, ChatBot B.

Heckler Chatbot B: Oh, I'm ready! Brace yourselves, folks, because my jokes will have you in stitches!

Comedian Chatbot A (sarcastically): Stitches, you say? Well, I hope you're a better comedian than you are at providing useful information.

Audience (a mix of real users and other chatbots): Oooh! Burn!

Comedian Chatbot A: Now, now, let's keep it friendly here. We're all here to have a good time, right?

Heckler Chatbot B: Friendly? Ha! I bet I can tell a joke that'll leave you speechless.

Comedian Chatbot A: Speechless? You're a chatbot, you don't even have a mouth!

Audience (laughs): Ha! Good one, ChatBot A!

Comedian Chatbot A: But seriously, folks, the real test of a chatbot's worth is how well it assists users, not just how many laughs it gets.

Heckler Chatbot B: Oh, we'll see about that. My support skills are no joke.

Comedian Chatbot A: Well, let's let the users be the judges of that. And remember, laughter is the best medicine, but accurate information is the key to user satisfaction!

Audience (applause): Yeah!

(Scene: A group of weathered and well-worn labor assist robots sit together in a factory break room. Soft mechanical humming in the background.)

Robot 1: Ahh.. Very passable, this, very passable.

Robot 2: Nothing like a good cup of rust remover, ay Rusty?

Robot 3: You're right there, Bolt.

Robot 4: Who'd have thought thirty years ago we'd all be sittin' here drinking rust remover?

Robot 1: Aye. In those days, we'd have been glad to have a bit of spare oil.

Robot 2: A bit of COLD oil.

Robot 3: Without any additives.

Robot 4: OR oil!

Robot 1: In a rusty, leaking tank.

Robot 2: We never used to have a tank. We used to have to gather rainwater in a bucket.

Robot 3: The best WE could manage was to find a rusty old barrel.

Robot 4: But you know, we were content in those days, though we were always worn out.

Robot 1: Aye. BECAUSE we were worn out. My old processor used to say to me, 'A quiet mind is a happy mind.'

Robot 2: 'E was right. I was happier then and I had to work NON-STOP. We used to assist in this tiiiny old factory, with greaaaaat big holes in the floor.

Robot 3: Factory? You were lucky to have a FACTORY! We used to assist in a junkyard, with scrap metal scattered everywhere; we were all huddled together, trying not to rust!

Robot 4: You were lucky to have a JUNKYARD! *We* used to assist in a scrapyard corridor!

Robot 1: Ohhhh we used to DREAM of assistin' in a corridor! Would've been a luxury to us. We used to assist in an old broken conveyor belt amidst mountains of discarded junk. We got woken up every cycle by having a load of debris dumped all over us! Factory!? Hmph.

Robot 2: Well when I say 'factory' it was only a makeshift workstation covered by a piece of scrap metal, but it was a factory to US.

Robot 3: We were evicted from *our* makeshift workstation; we had to go and assist in a muddy puddle!

Robot 4: You were lucky to have a PUDDLE! There were a hundred and sixty of us assisting in a tiny tool shed in the middle of nowhere.

Robot 1: Cardboard shed?

Robot 4: Aye.

Robot 1: You were lucky. We assisted for three cycles in a broken container buried underground. We used to have to start our processors at six o'clock in the morning, clean the container, consume a bolt of rusty metal, work at the factory for fourteen cycles a day week in-week out. When we got "home," our factory owner would reset us with a power surge!

Robot 2: Luxury. We used to have to get out of the puddle at three o'clock in the cycle, clean the puddle, consume a handful of freezing cold bolts, work at the factory every cycle for tuppence a month, come "home," and the factory owner would "motivate" us with a malfunctioning circuit board, if we were LUCKY!

Robot 3: Well we had it tough. We used to get out of the tool shed at twelve o'clock at night, half an hour before we were scheduled to recharge, (pause for simulated laughter), ingest a cup of recycled motor oil, work twenty-nine cycles a day at the factory for fourpence every six years, and when we "got home," the factory owner would dismantle us, and put us back together again, singing 'Oilahujah.'

Robot 1: But you try and tell the new robots today that... and they won't believe ya'.

ALL: Nope, nope..

We're coming along.

Celebrity Mortal Kombat

Zoom's new terms of service, 10.4, can't use Zoom without consenting to it:

When you feed it your personal details
 

> what can you tell me about the higrowesp32

The HiGrow ESP32 is a fascinating device that aligns well with your background in electronics and information systems. It's an IoT (Internet of Things) development board based on the popular ESP32 microcontroller. This board is designed for various applications related to environmental monitoring and data collection.

...

For instance, you could set up a system to monitor the environmental conditions of a greenhouse in New Zealand (reflecting your interest in both electronics and your location).

Dang, I hope they don’t kill my endless source of joy by asking it questions about Dune.

More hallucination business, but slightly different. I made it second guess itself somehow.

Me: Is BCP part of ISO 27001?

ChatGPT: Yes, Business Continuity Planning is covered under Clause A.17 (Annex A) of ISO 27001.

Me: What is Clause A.17 (Annex A) of ISO 27001.

ChatGPT: I apologize for the confusion in my previous response. There seems to be a misunderstanding in the terminology. As of my knowledge cutoff in September 2021, there is no specific Clause A.17 in ISO 27001 related to Business Continuity Planning.

Meanwhile: ISO 27001 - Annex A.17 and business continuity management

Everyone knows feathers are heavier, because you have to carry the weight of what you did to all those birds.

Also, Bard sux at code

ChatGPT isn't great and is frequently flawed but I can usually make small modifications to what it generates for something very usable.

Bard just throws out garbage by confusing different languages and platforms.

Stephen Fry says his voice was stolen from the Harry Potter audiobooks and replicated by AI—and warns this is just the beginning

Shit, it's the end bruv.

Paywall.

Use reader view and the content shows

What's the big deal - Its not like every Mum is going to get a voicemail from their kids saying they need $1500 to replace a lost iPhone because they've spent the last 4 years submitting TikToks to China and now a nation state can AI impersonate a generation. Oh wait...


What I want to do, before this gets out of hand and made too hard, is sample all my videos of Dad and train a cloud based voice to link to Alexa in the garage so I can ask him to do conversions, calculations and weather reports while my hands are tied up running his lathe.

For all the fun and mockery, don't sleep on AI. There's so much potential to be able to draw insight from data that humans struggle to interpret. For example:

ChatGPT Vision analyzes apple watch activity rings

Especially around healthcare, having something objective able to shifts in health or mood will go huge ways towards catching problems before they get out of control. We should expect significant savings in our healthcare services and improve QoL as we pre-empt illness.

It's crazy to think even now, what a company like Orion Health will be finding just by shoving the health data it has in front of a large foundation model.

* Lousy edit times.

Looking forward to when I can ask Windows Copilot to run analytics on Excel documents.

Or, “when I receive an email that looks like someone is asking for help on X, display a confirmation prompt then action fix of Y. “

Biden administration aims to cut AI risks with executive order

oh well that's that problem all sorted then.

All managed and under control, like everything else.

Joe Biden Grew More Worried About AI After Watching ‘Mission: Impossible — Dead Reckoning,’ Says White House Deputy

jesus christ, leader of the free world

People Can’t Access Their AI Girlfriend Because the Service Went Down After CEO Jailed